2 package com.hughes.android.dictionary.parser.wiktionary;
4 import com.hughes.android.dictionary.engine.HtmlEntry;
5 import com.hughes.android.dictionary.engine.IndexBuilder;
6 import com.hughes.android.dictionary.engine.IndexBuilder.TokenData;
7 import com.hughes.android.dictionary.engine.IndexedEntry;
8 import com.hughes.android.dictionary.parser.WikiTokenizer;
9 import com.hughes.util.StringUtil;
11 import org.apache.commons.lang3.StringEscapeUtils;
12 import org.apache.commons.lang3.StringUtils;
14 import java.util.ArrayList;
15 import java.util.LinkedHashMap;
16 import java.util.List;
18 import java.util.regex.Pattern;
20 public class WholeSectionToHtmlParser extends AbstractWiktionaryParser {
22 public static final String NAME = "WholeSectionToHtmlParser";
24 interface LangConfig {
25 boolean skipSection(final String name);
26 boolean skipWikiLink(final WikiTokenizer wikiTokenizer);
27 String adjustWikiLink(String wikiLinkDest);
28 void addFunctionCallbacks(
29 Map<String, FunctionCallback<WholeSectionToHtmlParser>> functionCallbacks);
31 static final Map<String,LangConfig> isoToLangConfig = new LinkedHashMap<String,LangConfig>();
33 final Pattern enSkipSections = Pattern.compile(".*Translations|Anagrams|References.*");
34 isoToLangConfig.put("EN", new LangConfig() {
36 public boolean skipSection(String headingText) {
37 return enSkipSections.matcher(headingText).matches();
41 public boolean skipWikiLink(WikiTokenizer wikiTokenizer) {
42 final String wikiText = wikiTokenizer.wikiLinkText();
43 if (wikiText.startsWith("Category:")) {
49 public String adjustWikiLink(String wikiLinkDest) {
50 if (wikiLinkDest.startsWith("w:") || wikiLinkDest.startsWith("Image:")) {
57 public void addFunctionCallbacks(
58 Map<String, FunctionCallback<WholeSectionToHtmlParser>> functionCallbacks) {
59 EnFunctionCallbacks.addGenericCallbacks(functionCallbacks);
63 final IndexBuilder titleIndexBuilder;
64 final String skipLangIso;
65 final LangConfig langConfig;
67 public WholeSectionToHtmlParser(final IndexBuilder titleIndexBuilder, final String wiktionaryIso, final String skipLangIso) {
68 this.titleIndexBuilder = titleIndexBuilder;
69 assert isoToLangConfig.containsKey(wiktionaryIso): wiktionaryIso;
70 this.langConfig = isoToLangConfig.get(wiktionaryIso);
71 this.skipLangIso = skipLangIso;
75 public void parseSection(String heading, String text) {
76 HtmlEntry htmlEntry = new HtmlEntry(entrySource, StringEscapeUtils.escapeHtml3(title));
77 IndexedEntry indexedEntry = new IndexedEntry(htmlEntry);
79 final AppendAndIndexWikiCallback<WholeSectionToHtmlParser> callback = new AppendCallback(
81 langConfig.addFunctionCallbacks(callback.functionCallbacks);
83 callback.builder = new StringBuilder();
84 callback.indexedEntry = indexedEntry;
85 callback.dispatch(text, null);
87 htmlEntry.html = callback.builder.toString();
88 indexedEntry.isValid = true;
90 final TokenData tokenData = titleIndexBuilder.getOrCreateTokenData(title);
92 htmlEntry.addToDictionary(titleIndexBuilder.index.dict);
93 tokenData.htmlEntries.add(htmlEntry);
94 // titleIndexBuilder.addEntryWithString(indexedEntry, title,
95 // EntryTypeName.WIKTIONARY_TITLE_MULTI_DETAIL);
99 void removeUselessArgs(Map<String, String> namedArgs) {
102 static final Pattern ALL_ASCII = Pattern.compile("[\\p{ASCII}]*");
104 class AppendCallback extends AppendAndIndexWikiCallback<WholeSectionToHtmlParser> {
105 public AppendCallback(WholeSectionToHtmlParser parser) {
110 public void onPlainText(String plainText) {
111 final String htmlEscaped = StringEscapeUtils.escapeHtml3(plainText);
112 if (ALL_ASCII.matcher(htmlEscaped).matches()) {
113 super.onPlainText(htmlEscaped);
115 super.onPlainText(StringUtil.escapeToPureHtmlUnicode(plainText));
120 public void onWikiLink(WikiTokenizer wikiTokenizer) {
121 if (wikiTokenizer.wikiLinkText().endsWith(":" + title)) {
122 // Skips wikilinks like: [[en::dick]]
125 if (langConfig.skipWikiLink(wikiTokenizer)) {
129 if (wikiTokenizer.wikiLinkDest() != null) {
130 linkDest = langConfig.adjustWikiLink(wikiTokenizer.wikiLinkDest());
132 linkDest = wikiTokenizer.wikiLinkText();
134 if (linkDest != null) {
135 builder.append(String.format("<a href=\"%s\">", linkDest));
136 super.onWikiLink(wikiTokenizer);
137 builder.append(String.format("</a>"));
139 super.onWikiLink(wikiTokenizer);
144 public void onFunction(WikiTokenizer wikiTokenizer, String name,
145 List<String> args, Map<String, String> namedArgs) {
146 if (skipLangIso.equalsIgnoreCase(namedArgs.get("lang"))) {
147 namedArgs.remove("lang");
149 super.onFunction(wikiTokenizer, name, args, namedArgs);
153 public void onHtml(WikiTokenizer wikiTokenizer) {
154 super.onHtml(wikiTokenizer);
158 public void onNewline(WikiTokenizer wikiTokenizer) {
162 public void onHeading(WikiTokenizer wikiTokenizer) {
163 final String headingText = wikiTokenizer.headingWikiText();
164 final int depth = wikiTokenizer.headingDepth();
165 if (langConfig.skipSection(headingText)) {
166 while ((wikiTokenizer = wikiTokenizer.nextToken()) != null) {
167 if (wikiTokenizer.isHeading() && wikiTokenizer.headingDepth() <= depth) {
168 wikiTokenizer.returnToLineStart();
174 builder.append(String.format("\n<h%d>", depth));
175 dispatch(headingText, null);
176 builder.append(String.format("</h%d>\n", depth));
179 final List<Character> listPrefixStack = new ArrayList<Character>();
182 public void onListItem(WikiTokenizer wikiTokenizer) {
183 if (builder.length() != 0 && builder.charAt(builder.length() - 1) != '\n') {
184 builder.append("\n");
186 final String prefix = wikiTokenizer.listItemPrefix();
187 while (listPrefixStack.size() < prefix.length()) {
188 builder.append(String.format("<%s>",
189 WikiTokenizer.getListTag(prefix.charAt(listPrefixStack.size()))));
190 listPrefixStack.add(prefix.charAt(listPrefixStack.size()));
192 builder.append("<li>");
193 dispatch(wikiTokenizer.listItemWikiText(), null);
194 builder.append("</li>\n");
196 WikiTokenizer nextToken = wikiTokenizer.nextToken();
197 boolean returnToLineStart = false;
198 if (nextToken != null && nextToken.isNewline()) {
199 nextToken = nextToken.nextToken();
200 returnToLineStart = true;
202 final String nextListHeader;
203 if (nextToken == null || !nextToken.isListItem()) {
206 nextListHeader = nextToken.listItemPrefix();
208 if (returnToLineStart) {
209 wikiTokenizer.returnToLineStart();
211 while (listPrefixStack.size() > nextListHeader.length()) {
212 final char prefixChar = listPrefixStack.remove(listPrefixStack.size() - 1);
213 builder.append(String.format("</%s>\n", WikiTokenizer.getListTag(prefixChar)));
217 boolean boldOn = false;
218 boolean italicOn = false;
221 public void onMarkup(WikiTokenizer wikiTokenizer) {
222 if ("'''".equals(wikiTokenizer.token())) {
224 builder.append("<b>");
226 builder.append("</b>");
229 } else if ("''".equals(wikiTokenizer.token())) {
231 builder.append("<em>");
233 builder.append("</em>");
235 italicOn = !italicOn;